Ticks and Tick-Borne Pathogens from Wild Pigs in Northern and Central Florida

Simple Summary Ticks are vectors of several agents of human disease, and improvements to traditional surveillance methods are needed to aid in tick-borne disease monitoring and prevention. Invasive wild pigs are broadly distributed in the southern U.S., and removal efforts are often undertaken by local, state, and federal entities. Wild pigs are hosts to several human-biting tick species associated with agents of human disease. This research examines the use of tick collection from wild pigs in the state of Florida as a method of surveillance for ticks and tick-borne pathogens of human concern. Four species of human-biting ticks were collected from wild pigs in this study, yielding similar results to traditional surveillance methods in the state. Known and potential human pathogens were identified in the collected ticks. Landscape features associated with tick diversity and abundance, such as developed spaces, mixed forest, and shrub/scrub habitat were identified, and may be useful for identifying areas of increased risk of encounters with human-biting ticks. These results help to inform tick and tick-borne pathogen surveillance efforts in the state of Florida and suggest that collections from wild pigs may be a useful surveillance tool for continued tick-borne disease surveillance. Abstract Invasive wild pigs are distributed across much of the U.S. and are hosts to tick vectors of human disease. Herein, adult ticks were collected from 157 wild pigs in 21 northern and central Florida counties from 2019–2020 during removal efforts by USDA-APHIS Wildlife Services personnel and evaluated for their potential to be used as a method of tick-borne disease surveillance. Collected ticks were identified, screened for pathogens, and the effects of landscape metrics on tick community composition and abundance were investigated. A total of 1415 adult ticks of four species were collected. The diversity of tick species collected from wild pigs was comparable to collections made throughout the state with conventional surveillance methods. All species collected have implications for pathogen transmission to humans and other animals. Ehrlichia, Anaplasma-like, and Rickettsia spp. were detected in ticks collected from wild pigs. These results suggest that tick collection from wild pigs is a suitable means of surveillance for pathogens and vectors. The strongest drivers of variation in tick community composition were the developed open space and mixed forest landcover classes. Fragmented shrub/scrub habitat was associated with increased tick abundance. Similar models may be useful in predicting tick abundance and distribution patterns.


Introduction
Pathogen surveillance generates data that can be used for outbreak prediction, prevention, and response by observation of the presence, prevalence, and changing distribution of pathogens and patterns of occurrence within populations [1]. The use of domestic or wild animals as surveillance sentinels can improve the efficiency of pathogen detection and play a role in the characterization of emerging diseases [1].
Modern sentinel surveillance employs a range of species and techniques. Surveys of seroprevalence, active infection, and mass or unexpected mortality in sentinel species can be used as an indicator of human risk or relative pathogen prevalence. For instance, bird die-offs have historically been used as an early warning signal preceding human cases of West Nile virus [2], and captive chickens are used as sentinel systems for the surveillance of several arboviruses [3]. The sero-surveillance of dogs was identified as an effective indicator of Borrelia burgdorferi presence in endemic areas [4][5][6]. Additionally, in areas of recent Ixodes scapularis Say range expansion in the U.S., dogs were determined to be effective sentinels of I. scapularis presence, while assays of both vector and non-vector tick species collected from dogs was useful in detecting the presence of B. burgdorferi [7]. These data suggest that, in addition to acting as sentinels for pathogen detection and surveillance, vertebrates also can be used for the surveillance of pathogen vectors.
The distributions of many tick species are expanding; the range of I. scapularis, the black-legged tick, is expanding both northward and westward [8][9][10], the range of Amblyomma americanum (L.), the lone star tick, is expanding northward [10,11], and the range of Amblyomma maculatum Koch, the Gulf Coast tick, is expanding northward and eastward [10,[12][13][14]. In 2017, Haemaphysalis longicornis Neumann, an invasive tick native to eastern Asia, was identified in New Jersey [15]. This tick was subsequently identified in 16 additional states on both domestic and wild animal hosts by the end of 2022 [16]. The collection of ticks from animal hosts may be an efficient method to monitor the spatial and temporal variation of vector tick species and pathogen prevalence, and to recognize the presence of introduced species.
Swine populations were introduced into what became the U.S. as early as the 1500s by Spanish explorers, and feral populations have been established following repeated introductions in the centuries following [17]. Modern introductions continue, with undomesticated wild Eurasian boar populations released in the 1900s for sport hunting continually being supplemented by free-roaming and escaped domestic pigs; the resulting intermixing feral populations have no natural predators, are highly adaptable, and are rapidly reproducing generalists [18,19]. The current range of wild pigs covers the south and southeastern U.S., California, and parts of several other western, northern and eastern states, with a total of 35 states having counties reporting a wild pig presence by 2021 [20].
Wild pigs may be hosts for several tick species, including A. americanum, A. maculatum, Dermacentor variabilis (Say), and I. scapularis [21], all of which are human-biting ticks known to transmit a variety of zoonotic pathogens. The use of tick sampling from wild pigs removed from public and private lands may be a useful method of surveillance for ticks and tick-borne pathogens of human concern.
Habitat types and landscape structures can affect the distribution of tick populations. Ostfeld et al. [22] found that questing I. scapularis abundance was greater in forested habitat types compared to shrubland or grassland; however, they found heavy larval infestation of rodents in shrubland and grassland habitats, despite low questing tick abundance. The authors suggest that host habitat use, driven by patch size and connectivity, largely determines the habitat associations of ticks. Some of the species that thrive in edge habitats at forest fragment boundaries, such as the white-footed mouse and white-tailed deer, are primary hosts for tick species such as I. scapularis and A. americanum [23][24][25]. The abundance of hosts utilizing fragmented habitat edges can result in an increase in tick abundance and an increase in the prevalence of associated pathogens; for example, studies have found positive correlations between I. scapularis density and increasing forest fragmentation [26,27]. Knowledge about the habitat variables associated with tick and pathogen abundance can aid in vector management, the prediction of human risk, and vector-borne disease prevention.
Given the expanding range of multiple tick species and the associated threat of tickborne disease, improvements to existing surveillance methods are imperative to mitigate the human health risk. The ability to predict and identify areas of increased tick abundance can aid in limiting human exposure to tick vectors. Due to the ecological destruction and the threat of disease transmission to domestic animals associated with wild pigs, removal efforts on the part of federal and state agencies and private landowners are frequently undertaken. The goal of this research was to determine if the collection of ticks from wild pigs can be used as a method to aid in tick and tick-borne pathogen surveillance efforts, and to determine if tick species abundance can be correlated with predictive environmental variables.

Tick Collection
Tick collection kits were distributed to United States Department of Agriculture Animal and Plant Health Inspection Service (USDA-APHIS) Wildlife Services employees who conducted swine live-trapping and euthanasia operations at approximately 30 trapping locations in the state of Florida. Kits contained a 2 mL BD Vacutainer ® EDTA tube (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) for whole blood collection from the swine via cardiac puncture, and two 1.5 mL screw-cap vials containing 95% ethanol for tick collection and preservation. Animal sampling procedures were approved under University of Florida Institutional Animal Care and Use Committee protocol 201910862.
From August 2019 to September 2020, adult and subadult wild pigs were sampled following euthanasia carried out by USDA-APHIS personnel. Personnel were instructed to collect up to 20 adult ticks per animal from a maximum of 10 pigs per location per season, targeting up to 10 adult ticks from the head and ears, and up to 10 adult ticks from the underline of the body, particularly the regions between the forelegs and the hind legs of the animal. The number of targeted ticks was limited to allow sampling to be completed in a timely manner while processing trapped animals. The GPS coordinates of the capture site were recorded with the collected ticks. Tick and blood samples were stored at −20 • C prior to transportation on ice to the Veterinary Entomology Laboratory at the University of Florida, where they were stored at −80 • C prior to DNA extraction. Samples were assigned to a season based on the sampling date. "Spring" samples were those collected from March-May (49 sampled pigs), "Summer" samples were those collected from June-August (37 sampled pigs), "Autumn" samples were those collected from September-November (41 sampled pigs), and "Winter" samples were those collected from December-February (30 sampled pigs).

Pathogen Screening
The collected ticks were taxonomically identified to life stage, sex, and species, and their engorgement status (flat vs. engorged) was recorded [28][29][30]. Only adult, flat ticks were used for DNA extraction and pathogen screening, to avoid host-blood inhibition of downstream reactions. A Gentra ® PureGene ® Blood kit (Qiagen, Valencia, CA, USA) was used for DNA extraction of laterally bisected ticks, following an optimized manufacturer protocol [31]. After Proteinase K digestion, the carapace of each tick was removed and placed in 75% ethanol to be held as a voucher specimen. Whole genomic DNA was extracted from whole pig blood using the same extraction kit following an optimized manufacturer-provided protocol. In short, the optimization included increased incubation and centrifugation times; doubling of the volume of the cell lysis solution, protein precipitation solution, and isopropoanol; the addition of Proteinase K (20 mg/mL concentration) for overnight sample incubation at 56 • C in a shaking incubator following the addition of cell lysis solution; and a repeat of the ethanol wash step to wash the DNA pellet.
Eluted samples were used to screen ticks for pathogens, including Anaplasma, Borrelia, Ehrlichia, and Rickettsia species. Conventional PCR was performed to screen for Anaplasma/Ehrlichia species using primers targeting the groEL gene followed by confirmation of the human or mouse strains of Anaplasma phagocytophilum positives using primers targeting the 16S rRNA gene [32,33]. Samples were screened for Borrelia species using primers targeting the Flagellin b gene [34], and spotted fever group Rickettsia species using primers targeting the ompA gene [35]. Swine blood samples were screened for infection with Anaplasma, Borrelia, and Ehrlichia using the same primer sets as for the tick screening to investigate the potential role of wild pigs in tick-borne pathogen cycles and to further demonstrate that any detected infections in ticks were representative of infective individuals and not the result of a blood meal from an infected host. Swine samples were not tested for Rickettsia infection due to the transovarial transmission patterns of tick-borne Rickettsia.
Amplicons from positive samples were submitted for bi-directional Sanger sequencing. Returned sequencing results were visualized and cleaned using GeneStudio™ software contig editor (GeneStudio, Inc., Suwanee, GA, USA). The resulting sequences were compared to sequences published in GenBank using a Basic Local Alignment Search Tool (BLAST) search to determine species identity. Confirmation of specimens identified with a low query coverage when compared to samples in GenBank was achieved by sequencing with an additional groEL primer set [36].

Statistical Analyses
To examine the effect of landscape variables on tick community composition, a 1 km radius buffer was generated around the trap location of each sampled animal using Ar-cMap™ v10.7.1 (ESRI ® , Redlands, CA, USA). A 1 km buffer size (3.14 km 2 area) was selected to encompass the likely area of use of wild pigs around the trapping site [17,37,38]. The 2019 National Landcover Raster was downloaded from the Multi-Resolution Land Characteristics Consortium (https://www.mrlc.gov/ Accessed: 12 November 2021). The raster was masked to the state of Florida in ArcMap. Five landcover classes hypothesized to most heavily affect wild pig use and tick abundance were selected for further analysis, and included: developed open space, deciduous forest, mixed forest, shrub/scrub, and herbaceous grassland. The edge density, patch size, and percent landcover of the five selected landcover classes were extracted from each 1 km buffer using the package "landscapemetrics" v1.5.4 [39]. A Pearson's correlation test was performed in R v1.2.5033 to identify highly correlated variables. Variables with a correlation coefficient < 0.7 and >−0.7 were included in the same models.
Partial redundancy analysis (pRDA) was performed using the "vegan" Community Ecology package v2.5-7 [40] to examine the effects of the three landscape features with the select landcover classes on the community composition of ticks from wild pigs. Tick count data were transformed using a Hellinger transformation [41]. Here, individual animals served as the site, and tick abundances served as species in the test. The Hellingertransformed animal-by-tick matrix served as the response variables. Latitude and longitude values provided location information on which to condition each of the models.
Three pRDA models were run, one for each of the landscape features. In the first model, the edge densities of the five landcover classes were the explanatory variables and the Hellinger-transformed animal-by-tick matrix served as the response variables. In the second model, the number of patches of each of the five landcover classes were the explanatory variables and the Hellinger-transformed animal-by-tick matrix served as the response variables. In the third model, the percent landcover of each of the five landcover classes were the explanatory variables, and the Hellinger-transformed animalby-tick matrix served as the response variables. A permutated ANOVA with 999 iterations was used to test the significance of explanatory variables.
We used generalized linear mixed effects models (GLMMs) to further investigate the effects of selected landscape variables and habitat types on A. americanum abundances. Pairs plots were generated to observe tick counts with each landcover variable to identify the presence of outliers and non-linearity. Because many landscape metrics are known to be highly collinear, variance inflation factor (VIF) values were calculated across all variables to test for collinearity in the dataset. High values were exhibited across the majority of variables. Thus, variables were divided into three candidate sets representing each of the landscape metrics investigated in the models (i.e., number of patches, edge density, and percent land cover) and the VIF calculations were repeated to ensure that multicollinearity was not present in environmental variables within each candidate set using a threshold VIF value < 3.0.
Visual inspection of the distribution of tick abundances revealed a large proportion of zeros and a wide distribution of counts along the x-axis, indicating strong potential for overdispersion [42]. In order to quantify the effects of edge density, number of patches, and percent landcover on A. americanum abundance, data were transformed using Log(x + 1), and fit to a GLMM with a Gaussian distribution including a pig-by-site random effect using the 'glmmTMB' package in R [43]. Models including all combinations of variables within each candidate set were generated using the "dredge" function in the 'MuMIn' package in R [44]. Results were evaluated using an information criterion approach, ranking models from lowest to highest Akaike's Information Criterion (AIC) scores, with the lower AIC scores indicating better-supported models [45]. In addition, AIC weights (AICw) were calculated for each model, providing information about the weight of evidence supporting the model, with higher values indicating stronger support. We used a threshold of delta AIC < 2 to identify the number of models contributing information. Effect curves for the most parsimonious model from each candidate set were generated with the 'ggeffects' package in R used to visualize the strength of the variable with 95% confidence intervals [46]. Model diagnostics of the most parsimonious model from each candidate set were performed using the 'Dharma' package in R [47], and model summaries include information about relevant predictors, those with CIs that did not cross zero, and p-values < 0.05. Tests for residual spatial autocorrelation were performed using a spatial correlogram in the 'ncf' package in R [48].  Figure 1). The damaged specimens were presumed to be Ixodes scapularis but were considered separately in further data analysis.  The greatest numbers of ticks were collected in the spring and summer (Table 1). Amblyomma americanum and D. variabilis were most frequently encountered in the spring and summer, A. maculatum in the autumn, and I. scapularis in the autumn and winter (Table 1). Ixodes scapularis and Ixodes species were more frequently encountered in the central portion of the state (Figure 1). Amblyomma maculatum was the least frequently encountered tick species but was distributed throughout the northern and central parts of the state ( Figure 1). Although A. americanum were the most frequently encountered ticks overall, they were absent at the southernmost collection sites (Figure 1).

Pathogen Screening
A total of 856 ticks from 20 counties (all unfed ticks collected) were screened for pathogens: 714 A. americanum, 18 A. maculatum, 75 D. variabilis, 37 I. scapularis, and 12 Ixodes sp. No ticks were infected with Borrelia species. There was a 3.3% infection prevalence of Ehrlichia and Anaplasma-like species (28/856) distributed among 19 A. americanum, 1 D. variabilis, 7 I. scapularis and one Ixodes sp. In total, 14 ticks were infected with E. ewingii: 13 A. americanum, >99% homologous to E. ewingii previously isolated from A. americanum (GenBank KJ907744), and one D. variabilis; the sequence quality of this amplicon was poor and was only 83.8% homologous to GenBank sequence KJ907744. Four A. americanum were infected with E. chaffeensis, >99% homologous to E. chaffeensis previously isolated from A. americanum (GenBank KJ907753), and two A. americanum were infected with Ehrlichia sp. Panola Mountain that was >99% similar to a previous isolate from A. americanum (GenBank KJ907753).
Seven I. scapularis and one Ixodes sp. were infected with a bacterial species identified with the groEL gene as A. phagocytophilum-like, having a low-percentage match to A. phagocytophilum (75.1-78.6% homologous to A. phagocytophilum GenBank MH722254); however, these samples were not amplified with the 16S primers targeting A. phagocytophilum. All eight samples were a >97% match to Candidatus Cryptoplasma californiense, identified from Ixodes pacificus Cooley and Kohls in California (GenBank KP276602), although the total query coverage of the sequences in GenBank was only 64%. The eight samples were subsequently amplified and sequenced using primers from Eshoo et al. [36] targeting the groEL gene. This increased the query cover to >95%, with six of the eight samples having >97% homology to Candidatus Cryptoplasma californiense identified from I. pacificus in California (GenBank KP276601). The remaining two of the eight samples were 87% and 91%, respectively, homologous to GenBank KP276601; these samples appeared to be co-infected, with sequences having multiple peaks.
Of the 383 Rickettsia-positive samples, 155 were sequenced. Rickettsia amblyommatis was detected in 137 A. americanum samples, 135 of which were >99% similar to R. amblyommatis amplicon sequences previously detected in an A. americanum collected in north-central Florida (GenBank MN313363). The two remaining A. americanum samples returned sequences that were a 97-98% match to the same isolate. One A. maculatum was infected with R. amblyommatis, with a sequence that was 98% similar to an R. amblyommatis isolate from A. americanum in Texas (GenBank MN336348). One A. maculatum was infected with R. parkeri with a sequence that was 100% similar to an R. parkeri isolate from A. maculatum in Texas (GenBank KP861344). One D. variabilis was infected with R. amblyommatis with a sequence that was 100% similar to GenBank MN313363. Three additional Rickettsia-positive D. variabilis samples that were collected from the same host animal as the sequenced sample were not sequenced. Ten I. scapularis and five Ixodes sp. were infected with a rickettsial endosymbiont that was ≥99% similar to an I. scapularis rickettsial endosymbiont detected in an I. scapularis collected in North Carolina (GenBank KP172259).
Bacteria species detected in ticks were distributed across collection sites in northern and central Florida but were not detected at the southernmost collection sites (Figure 2). No infection with Anaplasma, Borrelia, or Ehrlichia was detected in any swine blood samples (n = 157) from these sites.

Statistical Analyses
The results of the Pearson's correlation test indicated that there was little correlation among habitat types within each landscape feature (edge density, patch number, and percent landcover). The adjusted R 2 values of the proportion of constrained variance in the

Statistical Analyses
The results of the Pearson's correlation test indicated that there was little correlation among habitat types within each landscape feature (edge density, patch number, and percent landcover). The adjusted R 2 values of the proportion of constrained variance in the pRDAs in which tick community composition was the response variable and edge density, number of patches, and percent landcover of the five selected habitat types where the explanatory variables were 2.9%, 2.5%, and 5.5%, respectively. The proportion of variance conditioned on the X and Y coordinates of trap location was 12.8% in all three models ( Table 2). The models for edge density, patch number, and percent landcover were all significant (p = 0.014, p = 0.021, and p < 0.001, respectively).  When examining the relationship between A. americanum abundance and individual landscape metrics, the residual diagnostics showed slight non-linearity, but this factor was not strong enough to generate significant nonparametric dispersion. Variance inflation factor (VIF) values were low between all variables in each of the three candidate sets, indicating that there was little multicollinearity between the selected landcover classes within each landscape metric (Supplemental Material Table S1). Models of edge density and percent landcover of the selected landcover classes had the lowest AIC scores, while models of patch number had the highest AIC values (Table 3). Shrub/scrub and herbaceous grassland landcover classes were consistent variables in the best fit models.  When examining the relationship between A. americanum abundance and individual landscape metrics, the residual diagnostics showed slight non-linearity, but this factor was not strong enough to generate significant nonparametric dispersion. Variance inflation factor (VIF) values were low between all variables in each of the three candidate sets, Results for the edge density of the select landcover classes within a 1 km buffer distance as a predictor of A. americanum abundance on wild pigs indicated that five models were included in the best set of models, defined as models that do not exceed a threshold of ∆AIC < 2 (Table 3, Supplemental Material Table S2). In the best set of models of edge density, all five models included shrub/scrub and herbaceous grassland landcover classes as variables, followed by mixed forest and developed open space. Parameter estimates indicated a negative effect of herbaceous grassland, and a positive effect of shrub/scrub, mixed forest, and developed open space on A. americanum abundance. Deciduous forest was included as a variable in only one model, where parameter estimates indicated a negative effect on A. americanum abundance. The most parsimonious model had an AIC weight (AIC w ) of 0.205 and included edge density of shrub/scrub and herbaceous grassland as variables in the model. Both variables were significant ( Table 4). The effect curves for the edge density of shrub/scrub and herbaceous grassland indicated that, as values of edge density of shrub/scrub landcover increased, A. americanum abundance was predicted to increase; as values of herbaceous grassland increased, A. americanum abundance on wild pigs was predicted to decrease (Figure 4a).  The results for the number of patches of the select landcover classes measured within a 1 km buffer distance as a predictor of A. americanum abundance on wild pigs indicated that nine models were included in the best set of models (Table 3, Supplemental Material  Table S2). All landscape classes appeared as variables multiple times within these nine models, although shrub/scrub was the most frequently included variable, followed by mixed forest. Parameter estimates indicated a positive effect of both variables on A. americanum abundance. The most parsimonious model had an AIC w of 0.106 and included only number of patches of shrub/scrub as a variable in the model, and this variable was significant ( Table 4). The effect curve for the number of patches of shrub/scrub landcover indicated that as the number of patches of shrub/scrub landcover increased, A. americanum abundance on wild pigs was predicted to increase (Figure 4b).
The results for the percent landcover of the select landcover classes measured within a 1 km buffer distance as a predictor of A. americanum abundance on wild pigs indicated that three models were included in the best set of models (Table 3, Supplemental Material  Table S2). The percent landcover of shrub/scrub and herbaceous grassland were included as variables in all three models. Percent landcover of mixed forest and of developed open space were each included as an additional variable in one of the three models. Parameter estimates indicated a positive effect of the percent landcover of shrub/scrub on A. americanum abundance, and a negative effect of the percent landcover of herbaceous grassland on A. americanum abundance. The most parsimonious model included percent landcover of shrub/scrub and percent landcover of herbaceous grassland as variables in the model. A polynomial term was added to the model covariates to improve the model fit, wherein shrub/scrub was a significant variable, while herbaceous grassland was marginally significant ( Table 4). The effect curves for the percent landcover of shrub/scrub and herbaceous grassland indicated that, as the percent of shrub/scrub landcover increased, A. americanum abundance was predicted to increase, and that as the percent landcover of herbaceous grassland increased, A. americanum abundance was predicted to decrease ( Figure 4c).
The model diagnostics indicated a good model fit for all models described (Supplemental Material Figure S1). Tests for residual spatial autocorrelation were performed using a spatial spline correlogram in the 'ncf' package in R. The results indicated that significant autocorrelation was not present in the most parsimonious model of each of the landscape metrics (Supplemental Material Figure S2). Insects 2023, 14, x FOR PEER REVIEW 13 The results for the number of patches of the select landcover classes measured w a 1 km buffer distance as a predictor of A. americanum abundance on wild pigs ind that nine models were included in the best set of models (Table 3, Supplemental Ma  Table S2). All landscape classes appeared as variables multiple times within these models, although shrub/scrub was the most frequently included variable, follow

Discussion
The diversity of tick species collected from the wild pigs was comparable to prior collections made throughout the state with conventional surveillance methods such as dragging and flagging [49,50], and sampling from animals [51][52][53][54]. All the tick species collected have implications for zoonotic pathogen transmission. Thus, collections from wild pigs can be a useful method to aid in statewide surveillance for tick pathogen vectors. While the species of ticks detected in this study are comparable to other large-scale surveys in the state, the relative abundances of each of the detected species differed from some studies. Proportionately more A. maculatum and D. variabilis were collected from wild pigs in this study than from surveys of questing ticks in the state, while proportionately fewer I. scapularis were detected [49,50]. Allan et al. [51] reported that I. scapularis was the most common tick they collected from wild pigs, followed by A. americanum. Their sampling, however, was carried out exclusively during the autumn deer hunting season, which coincides with the peak activity of adult I. scapularis and could account for the increased abundance of I. scapularis. Hertz et al. [53] sampled a variety of wildlife host species but detected similar relative abundances of the four species detected herein; Hertz et al. also identified small numbers of two species (Ixodes affinis Neumann and Ixodes texanus Banks) that were not detected from the swine herein. A southern Florida study that collected ticks from wild pigs, and by dragging on the same property, detected four species of adult ticks from wild pigs and only three species from drags, although immature-stage ticks were more frequently collected on drags [54]. Notably, tick species that are less frequently collected by dragging/flagging, such as A. maculatum and D. variabilis, may be encountered more frequently when sampling from host animals; therefore, animal sampling may be a useful method for monitoring the presence of ticks that are underrepresented by other collection methods.
Although immature ticks may be less frequently detected on wild pigs, adult ticks, which have completed multiple blood meals, are often the preferred stage for general pathogen surveillance due to transstadial accumulation of pathogens. In this study, collectors were instructed to collect only adult ticks to maximize sampling efforts.
Three pathogenic species of Ehrlichia were detected in ticks collected from wild pigs. Ehrlichia chaffeensis, E. ewingii, and Ehrlichia sp. Panola Mountain each have been previously detected, typically at low prevalence, in questing and host-collected ticks in Florida [49,[55][56][57]. The overall prevalence of E. chaffeensis in this study was 0.47%, the prevalence of E. ewingii was 1.6%, and the prevalence of Ehrlichia sp. Panola Mountain was 0.23%. The 20 ticks infected with pathogenic Ehrlichia came from 14 animals at nine locations in seven counties ( Figure 2). The Ehrlichia prevalence detected herein is similar to most detections in prior studies from both questing and host-collected ticks. This indicates that tick collections from wild pigs as a surveillance method for vector-borne pathogens yields similar estimates of pathogen prevalence as dragging and flagging.
Seven I. scapularis and one Ixodes sp. were infected with an A. phagocytophilum-like organism, Candidatus Cryptoplasma californiense, that was originally identified in I. pacificus in California. In their initial description of this organism, Eshoo et al. [36] noted the detection of isolates with high similarity to their California isolates from studies in eastern Asia, Europe, and northern Africa, and suggested that this organism may be widespread in Ixodes. The potential for this Cryptoplasma species to cause illness in humans or other animals is unknown. Five of the eight infected ticks were removed from the same animal. The other three infected ticks came from three different animals. All four animals were trapped at the same collection site in the northeastern corner of the state (Figure 2). To our knowledge, this was the first detection of this organism in ticks in Florida.
The high prevalence of Rickettsia in ticks collected from wild pigs was driven by the 73% prevalence of rickettsial endosymbionts in Ixodes ticks and the 48% prevalence of Rickettsia species in A. americanum. All sequenced amplicons from A. americanum were identified as R. amblyommatis. Infection with R. amblyommatis also was detected in one D. variabilis and one A. maculatum. It is notable that R. parkeri-infection was detected in only one of 18 A. maculatum samples, as a high prevalence of Rickettsia spp. (~30%), primarily identified as R. parkeri in A. maculatum from Florida populations, has previously been reported [56]. With the exception of R. parkeri, the Rickettsia species detected in this study were either non-pathogenic or of questionable pathogenicity [57].
Landscape fragmentation and the availability of edge habitat was shown to be an important driver of the presence of various tick species, likely due to the close association between several tick species and host species that utilize edge habitat [22][23][24][25][26][27]. In this study, the redundancy analysis suggested that the edge density and patch number of mixed forest and developed open space, and the percent landcover of mixed forest, developed open space, and grassland landcover were the major drivers of variation in tick community composition (Figure 3). The collected tick species were most closely associated with deciduous forest, shrub/scrub, herbaceous grassland and open developed space. As landscape fragmentation increases, tick community composition is also expected to be affected.
The data collected in this study were used to produce GLMMs to determine which landscape metrics were most strongly correlated with the abundance of A. americanum, the most commonly encountered tick in the study. Two of the utilized metrics, edge density and number of patches of selected landcover classes, served as indicators of increased fragmentation. The most parsimonious models within each landscape metric of A. americanum abundance always included the shrub/scrub landcover class (Table 3, Supplemental Material Table S2). As the edge density, number of patches, or percent landcover of the shrub/scrub landcover class increased, the predicted values of A. americanum abundance increased as well. The most parsimonious model for two of the three landscape metrics also included the herbaceous grassland landcover type as a covariable. As the edge density or percent landcover of herbaceous grassland increased, the predicted values for A. americanum abundance decreased.
The strongest models generated herein with the lowest AIC values and highest AIC w , or relative likelihood, were the model of edge density where shrub/scrub and herbaceous grassland were included as covariables, and the model of percent landcover where shrub/scrub and herbaceous grassland as covariables. These results demonstrate that these two landcover types are important drivers of tick abundance. The herbaceous grassland landcover likely has a direct negative effect on A. americanum abundance, while species such as A. maculatum seem to thrive in grassland habitat [58,59]. The positive effect of shrub/scrub habitat may be the result of host use and movement patterns, such as the use of the shrub/scrub habitat by wild pigs as they move between areas of resource utilization (i.e., foraging near the edges of pasture, cropland, or grassland, but sheltering in more wooded areas) and utilization by generalist species such as white-tailed deer, another primary host of A. americanum [25]. A Florida study that evaluated environmental associations to predict A. americanum distribution found that their forest variable was among those predicted to most heavily influence A. americanum presence [60]. However, this study surveyed questing ticks. Other studies have indicated that the abundance of ticks on a host in an area may be high even when questing tick numbers are low due to host movement patterns [22].
The collection of four human-biting tick species from wild pigs suggest that this host sampling method is promising for human tick-borne pathogen and vector surveillance. Our sampling of wild pigs resulted in the collection of a greater abundance of tick species that can be difficult to detect with traditional sampling methods such as flagging/dragging; thus, this surveillance method may be useful in detecting the presence or monitoring the range expansion of species that may otherwise be difficult to detect. The identification of human pathogens in ticks collected from wild pigs, and the detection of a novel tick-borne bacteria in Florida from wild pig-collected ticks further demonstrate the usefulness of wild pigs as a method of surveillance for ticks and tick-borne pathogens. The integration of large-scale sampling from wild pigs with other surveillance methods can augment and improve current vector-borne disease surveillance efforts.
Supplementary Materials: The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/insects14070612/s1, Figure S1: QQ plots, residual diagnostics, and dispersion tests for model fit of most parsimonious model of each candidate set used to predict abundance of Amblyomma americanum collected from 157 wild pigs in north and central Florida. (a) Edge density, (b) Number of patches, (c) Percent landcover; Figure S2: Spatial correlogram plot with 95% confidence intervals to test for residual spatial autocorrelation in each of the three landscape variables used to predict abundance of A. americanum collected from wild pigs in north and central Florida. 1000 iterations were used to generate a bootstrap distribution; Table S1: Variance Inflation Factor (VIF) values for each candidate set of variables within each landscape metric for the prediction of the abundance of A. americanum from wild pigs in northern and central Florida; Table S2: Akaike Information Criterion (AIC) table of models of three landscape metrics where selected landcover classes are predictor variables to predict the abundance of A. americanum collected from wild pigs in north and central Florida (2020-2021).